The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external file content, creating an indirect prompt injection surface.
Ingestion points: Reads content from local SKILL.md and README.md files in Phase 1.
Boundary markers: Content is placed within structured MDX headers such as ## What it does.
Capability inventory: The skill can write files, perform rsync operations, and execute git push.
Sanitization: External content is copied verbatim or truncated without specific sanitization for MDX or shell characters.
[COMMAND_EXECUTION]: The skill utilizes several shell commands to manage files and repositories.
Runs rsync to synchronize content between the internal Claude skills directory and the project repository.
Uses git for commit history logging, staging changes, and performing commits.
Uses date for metadata generation.
[DATA_EXFILTRATION]: The skill transmits local project data to remote repositories via git push. While this is the intended functionality for publishing, it involves moving local content to a remote server.

publish-skill